A Randomized Experiment of Offsite Compliance Monitoring Activities in Environmental Regulations

How to cite

Andarge, Tihitina; Evans, Mary; Ferraro, Paul. 2025. "A Randomized Experiment of Offsite Compliance Monitoring Activities in Environmental Regulations." AEA RCT Registry. AEARCTR-0017552. https://doi.org/10.1257/rct.17552-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Tihitina Andarge* , Mary Evans , Paul Ferraro

* Primary investigator

Registry ID
AEARCTR-0017552
DOI
https://doi.org/10.1257/rct.17552-1.0
First registered
Last updated
Status
In development
Trial dates
2026-01-01 – 2026-12-31
Countries
Private
JEL codes
Q50, Q58, K32
Registry record
http://www.socialscienceregistry.org/trials/17552

Abstract

In partnership with the Wisconsin Department of Natural Resources (DNR) and the Office of Enforcement and Compliance Assurance (OECA) at the U.S. Environmental Protection Agency (EPA), we aim to conduct a pilot experiment to compare offsite desk audits and onsite inspections within the context of the Clean Water Act's National Pollution Discharge Elimination System (NPDES) wastewater program in Wisconsin. Within inspector-half year blocks, we will randomly assign facilities to receive an offsite desk audit (treatment group) or an onsite inspection (control group). For decision-making, states are interested in knowing if offsite compliance monitoring (OfCM) activities are no worse than onsite inspections in their ability to uncover violations, rather than whether they are better than onsite inspections. This setting lends itself to a non-inferiority randomized controlled trial (RCT) where we test whether offsite desk audits are no worse than onsite inspections by some predetermined margin of acceptable difference, referred to as the noninferiority margin. The noninferiority margin was elicited from our state partner. Over the course of the experiment, inspectors will fill out a standard inspection form for both groups of facilities. At the conclusion of the experiment, we will extract the data from the forms to determine the number of violations, inspection date, inspector, and monitoring type. We will then perform a noninferiority test on the mean number of violations in the two groups. As secondary outcomes, we will also examine the time spent on monitoring and the cost of monitoring.

Experimental design

From our state partner, we obtained a list of facilities that meet our inclusion criteria and are due for an inspection between January 1 and August 31, 2026. Each facility already has a designated inspector. Within inspector-half-year blocks, we randomly assign facilities to be monitored via an offsite desk audit or an onsite inspection. For the first half of 2026, between January 1 and June 30, 2026, we performed the randomization for the 66 eligible facilities due for an inspection during that time period. We provided a spreadsheet with the treatment assignments to our state partner. For the second half of 2026, we have to wait until April 2026 to perform the randomization, once our state partner has constructed the full list of eligible facilities for that time period. We expect an additional 60 facilities to be included in the second half of 2026, bringing the total expected number of observations to 126 facilities.

Design details

Randomization method
Randomization was performed using the function "block_ra" from the "randomizeR" package in R.
Randomization unit
Within inspector-half-year blocks, facilities were randomly assigned to receive an offsite desk audit (treatment group) or an onsite inspection (control group).
Clusters
N/A
Observations
We expect about 126 facilities.
Arms
We expect 63 facilities in the treatment group (offsite desk audits) and 63 facilities in the control group (onsite inspections).
Primary outcomes
The primary outcome variable is the number of violations.
Secondary outcomes
The secondary outcome variables include the time spent on monitoring and the cost of monitoring.

Cited by

No citation data available for this registration yet.

Keywords